Research map / Probabilistic and scientific modeling / Variational and generative sampling
Self-Supervised Latent Space Optimization With Nebula Variational Coding
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add the mass term M_a to the nebula loss (vs Euclidean distance to anchor) (ablation) | NVC on ForkNet · ShapeNet real completion | IoU (%): 24.9 -> 22.1 without M_a | Improved |
| add the mass term M_a to the nebula loss (ablation) | NVC on ForkNet · CompleteScanNet | IoU (%): 14.1 -> 11.5 without M_a | Improved |
| add L_pair (Siamese) to the metric learning (ablation) | NVC-ML on ForkNet · ShapeNet real completion | IoU (%): 26.3 -> 24.5 without L_pair | Improved |
| add L_triplet to the metric learning (ablation) | NVC-ML on ForkNet · ShapeNet real completion | IoU (%): 26.3 -> 23.9 without L_triplet | Improved |
| add L_triplet to the metric learning (ablation) | NVC-ML on ScanComplete · CompleteScanNet | IoU (%): 19.8 -> 17.3 without L_triplet | Improved |
| add self-supervised metric learning on top of NVC | 3D-RecGAN · ShapeNet real completion | IoU (%): 19.6 -> 20.5 | Improved |
| replace GMVAE clustering with NVC | NMT-GNMT · WMT13 De-En | BLEU: 29.8 -> 31.2 | Improved |
| add the mass term M_a with many anchors (ablation) | NVC with 3D-RecGAN · ShapeNet completion, 40 anchors | IoU (%): 74.0 -> 79.3 | Improved |
| add NVC to VO-Hand on unoccluded data | VO-Hand (cluster) · Stereo hand pose | AUC: 0.984 -> 0.986 | No clear effect |
| add NVC to VO-Hand (triplet) on occluded data | VO-Hand · HOP hand pose | AUC: 0.597 -> 0.623 | Improved |
| add nebula anchors to the latent space | PlanarRecon · NYUv2 | rel error: 0.134 -> 0.126 | Improved |
| add 10 anchors with metric learning | NVC-ML on MNIST autoencoder · MNIST reconstruction | rel error: 0.169 -> 0.133 | Improved |
Closest papers by meaning
- Learning Latent Superstructures in Variational Autoencoders for Deep Multidimensional Clustering (ICLR 2019 · similarity 0.88)
- Mixed-curvature Variational Autoencoders (ICLR 2020 · similarity 0.87)
- Disentangling Latent Space for VAE by Label Relevant/Irrelevant Dimensions (CVPR 2019 · similarity 0.87)
- Multi-Facet Clustering Variational Autoencoders (NeurIPS 2021 · similarity 0.86)
- MAE: Mutual Posterior-Divergence Regularization for Variational AutoEncoders (ICLR 2019 · similarity 0.86)
- Shape your Space: A Gaussian Mixture Regularization Approach to Deterministic Autoencoders (NeurIPS 2021 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
